With respect to decal manipulation to make NCC-1017, for the sake of completeness, it's worth pointing out that there are
two ways to make 1017 from 1701 with the same degree of minimal fuss, by only two snips:
1701 >> 17 0 1 >> 1 0 17 >> 1017
1701 >> 1 7 01 >> 1 01 7 >> 1017
Also, with respect to the question of why 1017, if we consider all possible permutations using a single decal, there are some interesting results. If we reject a leading 0 as starting a number that's too low and/or one that should have been three digits to begin with, and we reject a leading 7 as starting a number that's too high, then there are only five possibilities (not counting 1701).
1017
1071
1107
1170
1710
Of these five, there are only two that, besides the leading 1, have no digit in the same position as 1701. These two would be the ones that are arguably most visually distinct from 1701. These two are:
1017
1170
Both of these can be made with two snips of the decal in two different ways (three in fact, in the case of 1170, if you count that in the first example below, there are two ways to order the 1s). The ways for 1017 were shown above. For 1170, we have:
1701 >> 1 70 1 >> 1 1 70 >> 1170
1701 >> 17 0 1 >> 1 17 0 >> 1170
The last example has the same snip pattern as the first example for making 1017, it's just that the final arrangement is different.
I'm not saying that making the
Constellation's registry number as distinct as possible from the
Enterprise's is why 1017 was chosen, but in my opinion it is an interesting observation nonetheless. The only outstanding question would be, then, why 1017 and not 1170? Both appear to be equally good candidates. Perhaps the choice was simply random.
But if you need an additional reason, one is in fact available, and that is that 1017 is more visually distinguishable on the left from 1701, because at the second position, 0 looks less like 7 than 1 does.
And there we have it. The use of these two principles would have made the choice unique, 1) making sure none of the free digits are in the same position as they are in 1701, so that no matter where you look (after the first digit at least, which as argued must be the same) you will see a different digit and 2) maximizing the distinction on the left, so that when the number is read left to right you can see the distinction most easily in the least amount of time.
Of course, none of this means that it's why the selection was made.
edited to add - This post did not take into account the possibility of effectively rotating the decals as described down-thread in post #364:
https://www.trekbbs.com/threads/that-starbase-11-wall-chart-noe-in-slide-form.319240/post-15269618
See also post #366 regarding the curved baseline that the decals apparently had in the original kit:
https://www.trekbbs.com/threads/that-starbase-11-wall-chart-noe-in-slide-form.319240/post-15269641